ctdb reclock API for etcd/librados

[José A. Rivera]

On 2016-12-09 06:40, David Disseldorp wrote:
> Hi Martin,
> 
> On Fri, 9 Dec 2016 21:13:52 +1100, Martin Schwenke wrote:
> 
>> Hi David,
>> 
>> On Thu, 8 Dec 2016 15:13:42 +0100, David Disseldorp <ddiss at suse.de>
>> wrote:
>> 
>> > On Thu, 8 Dec 2016 17:42:14 +1100, Amitay Isaacs wrote:
>> >
>> > ...
>> > > Due to the unconditional addition of ctdb-recd-lock, our nightly CTDB
>> > > builds/tests are failing.  That's the reason for urgency (and my bad mood).
>> >
>> > Okay, thanks for the explanation. Out of interest, how was the new
>> > executable causing the build/test failure? Can we get the same checks
>> > into autobuild, to avoid this in future.
>> 
>> The extra files cause RPM build failures.
>> 
>> We build RPMs to install on the nodes of the virtual clusters that we
>> build to test with.  This works well and only really matters because 
>> we
>> need a consistent way of installing across a cluster - not something
>> that is needed with non-clustered Samba.
>> 
>> We could do:
>> 
>>   %define _unpackaged_files_terminate_build 0
>> 
>> but then we could miss new core files that we definitely want to
>> package.  This happens to me occasionally when I forget to update the
>> RPM spec file.
> 
> Ah okay, yeah that's a pretty common packaging challenge :-)
> I guess we could make an effort to keep ctdb/packaging/RPM/ctdb.spec.in
> up to date with any changes to the installation file manifest.
> 
> Cheers, David

Do we still need to update the spec file now for these new files?

--Jose